Measuring gender differences in partner violence: implications from research on other forms of violent and socially undesirable behavior

Sex Roles: A Journal of Research,  June, 2005  by Sherry L. Hamby

• E-mail
• Print
• Link

Difficulties With Referent Periods

It is challenging to remember events that occurred during an arbitrary period of time, such as the 12 months preceding a survey. There are two major types of memory bias: memory fading for events in the more distant past, and telescoping, which is a tendency to report events as occurring at a different time than they actually did. Memory fading and backward telescoping (identifying events as occurring before the referent period when they occurred within it) produce false negatives. Forward telescoping produces false positives because events that did not actually occur in the recall period are reported as if they were.

Technical studies done during the development phase of the NCVS/NCS have provided the most conclusive evidence of the extent of these problems. In one reverse-records check study, 69% of events from the previous 90 days were reported versus only 30% for the 10-12 months prior to interview (Turner, 1972). In another, many more incidents were reported as occurring within the last 6 months rather than the first 6 months of a calendar year (U.S. Department of Justice, 1974), which could have been due to memory fading or forward telescoping or both. Either way, these data suggest that memory problems might lower 12-month incident rates. Although the fairly low base rate of partner violence is one reason for longer referent periods (Campbell, 2000; Straus et al., 1996), memory difficulties still cloud the interpretation of obtained rates.

Reporting Load

"Reporting load" refers to the problem of a respondent attempting to shorten an interview by omitting relevant incidents (Czaja et al., 1994). Reporting load cautions against simply lengthening an interview to increase disclosures. For example, reporting load may have affected responses to the NVAWS. This was an admirable effort to obtain comprehensive violence data, including questions on rape, psychological aggression, stalking, and physical assault (Tjaden & Thoennes, 1998). The resulting survey included dozens of questions, many very sensitive, on violence. This appears to have led to respondent fatigue and may be one reason why the physical assault rates, which appeared at the end of the survey, are so out of line with most other research using similar methodology. For example, a single question in a 1995 nationally representative Gallup survey (Gallup Organization, 1995) produced a rate that was three times higher than the NVAWS rate even though single items almost certainly lead to underreporting. A Commonwealth Fund survey conducted after the NVAWS produced a rate almost four times higher (Collins, Schoen, Joseph, Duchon, Simantov, & Yellowitz, 1999). The NVAWS rate suggests an approximately 80% drop in partner violence rates since the time of the NFVS surveys, which seems unlikely. The NVAWS rate also suggests the US rate is lower than rates in Canada (Johnson & Sacco, 1995), Switzerland (Gillioz, De Puy, & Ducret, 1997), and other countries with less violent crime. Further, the NVAWS, like the NCVS, tried to assess violence by all possible perpetrators, which may also add to reporting load and be one reason why both surveys produce low rates. Certainly, researchers should be cautioned about creating reporting load.